Approximation algorithms employing Monte Carlo methods, across application domains, often require as a subroutine the estimation of the mean of a random variable with support on [0,1]. One wishes to estimate this mean to within a user-specified error, using as few samples from the simulated distribution as possible. In the case that the mean being estimated is small, one is then interested in controlling the relative error of the estimate. We introduce a new (epsilon, delta) relative error approximation scheme for [0,1] random variables and provide a comparison of this algorithm's performance to that of an existing approximation scheme, both establishing theoretical bounds on the expected number of samples required by the two algorithms and empirically comparing the samples used when the algorithms are employed for a particular application.

In this thesis, a discrete Markov chain model for analyzing memory interference in multiprocessors, is presented.

Mathematical modeling is a central component of operations research. Most of the academic research in our field focuses on developing algorithmic tools for solving various mathematical problems arising from our models. However, our procedure for selecting the best model to use in any particular application is ad hoc. This dissertation seeks to rigorously quantify the trade-offs between various design criteria in model construction through a series of case studies. The hope is that a better understanding of the pros and cons of different models (for the same application) can guide and improve the model selection process. In this dissertation, we focus on two broad types of trade-offs. The first type arises naturally in mechanism or market design, a discipline that focuses on developing optimization models for complex multi-agent systems. Such systems may require satisfying multiple objectives that are potentially in conflict with one another. Hence, finding a solution that simultaneously satisfies several design requirements is challenging. The second type addresses the dynamics between model complexity and computational tractability in the context of approximation algorithms for some discrete optimization problems. The need to study this type of trade-offs is motivated by certain industry problems where the goal is to obtain the best solution within a reasonable time frame. Hence, being able to quantify and compare the degree of sub-optimality of the solution obtained under different models is helpful. Chapters 2-5 of the dissertation focus on trade-offs of the first type and Chapters 6-7 the second type.

本论文着重研究了带有Lp模球约束以及二次约束的多项式优化问题的计算复杂度以及关于此类问题的近似算法。在本论文中，利用张量对称化的技巧，我们首次证明了当P∈ [2 ，∞] ，任意高阶的带有Lp模球约束的多项式优化问题均为NP 困难。借助模的对偶性质，我们将这类优化问题转化为求解凸体半径的问题，从而使得我们获得了之前研究所无法使用的算法工具。具体来说，利用计算凸几何的算法工具，对于Lp模球约束的多项式优化问题，我们得到了近似比为[附圖]的确定性多项式时间近似算法，其中d为目标多项式的阶次， n 为问题的维度。使用随机算法，我们将近似比进一步提高为此类问题的己知最优值。[附圖]。此外，我们发展了计算凸几何当中对于凸体半径的计算方法，从而设计出了一种对二次约束多项式优化问题近似比为[附圖]的近似算法，其中m为问题的约束个数。我们的结果涵盖并提高了之前关于此类问题的研究结果。我们相信在本论文中使用的新的算法工具，将在今后的多项式优化问题研究中得到更广泛的应用。 / In this thesis, we present polynomial time approximation algorithms for solving various homogeneous polynomial optimization problems and their multilinear relaxations. Specifically, for the problems with Lp ball constraint, where P∈ [2 ，∞], by reducing them to that of determining the Lq-diameter of certain convex body, we show that they can be approximated to within a factor of [with formula] in deterministic polynomial time, where q = p=(p - 1) is the conjugate of p, n is the number of variables, and d is the degree of the polynomial. We further show that with the help of randomization, the approximation guarantee can be improved to [with formula], which is independent of p and is currently the best for the aforementioned problems. Moreover, we extend the argument of deterministic algorithm mentioned above to solve the quadratically constrained polynomial optimization problems. In particular, for any intersection of ellipsoids K, we can, in polynomial time, construct a random polytope P, which satisfies [with formula]. Then, by reducing the problem to that of evaluating the maximum polytopal norm [with formula] induced by P, over certain convex body, we can approximate the quadratically constrained problem within a factor of [with formula] in polynomial time. Our results unify and generalize those in the literature, which focus either on the quadratic case or the case where [with formula]. We believe that the wide array of tools used in this thesis will have further applications in the study of polynomial optimization problems. / Detailed summary in vernacular field only. / Hou, Ke. / On title page "p" is subscript. / Thesis (Ph.D.) Chinese University of Hong Kong, 2013. / Includes bibliographical references (leaves 106-111). / Abstracts also in Chinese.

Online Bin Stretching: Algorithms and Computer Lower Bounds Author: Martin Böhm Abstract: We investigate a problem in semi-online algorithm design, called Online Bin Stretching. The problem can be understood as an online repacking problem: the goal of the algorithm is to repack items of various sizes into m containers of identical size R > 1. The input items arrive one by one and the algorithm must assign an item to a container before the next item arrives. A specialty of this problem is that there is a specific guarantee made to the algorithm: the algorithm learns at the start of the input that there exists a packing of all input items into m containers of capacity 1. Our goal is to design algorithms for this problem which successfully pack the entire incoming sequence one by one while requiring the lowest container capacity R possible. In this thesis, we show several new results about Online Bin Stretching: First, we design an algorithm that is able to pack the entire input into m containers of capacity 1.5 regardless of what the vale of m will be. Second, we show a specialized algorithm for the setting of just 3 containers; this algorithm is able to pack into 3 bins of capacity 1.375. Finally, we design and implement an involved search algorithm which is able to find lower bounds for Online Bin...

We consider several combinatorial optimization problems in a geometric set- ting. The first problem we consider is the problem of clustering to minimize the sum of radii. Given a positive integer k and a set of points with interpoint distances that satisfy the definition of being a "metric", we define a ball centered at some input point and having some radius as the set of all input points that are at a distance smaller than the radius of the ball from its center. We want to cover all input points using at most k balls so that the sum of the radii of the balls chosen is minimized. We show that when the points lie in some Euclidean space and the distance measure is the standard Euclidean metric, we can find an exact solution in polynomial time under standard assumptions about the model of computation. The second problem we consider is the Network Spanner Topology Design problem. In this problem, given a set of nodes in the network, represented by points in some geometric setting - either a plane or a 1.5-D terrain, we want to compute a height assignment function h that assigns a height to a tower at every node such that the set of pairs of nodes that can form a direct link with each other under this height function forms a connected spanner. A pair of nodes can form a direct link if they are within a bounded distance B of each other and the heights of towers at the two nodes are sufficient to achieve Line-of-Sight connectivity - i.e. the straight line connecting the top of the towers lies above any obstacles. In the planar setting where the obstacles are modeled as having a certain maximum height and minimum clearance distance, we give a constant factor approximation algorithm. In the case where the points lie on a 1.5-D terrain we illustrate that it might be hard to use Computational Geometry to achieve efficient approximations. The final problem we consider is the Multiway Barrier Cut problem. Here, given a set of points in the plane and a set of unit disk sensors also in the plane such that any path in the plane between any pair of input points hits at least one of the given sensor disks we consider the problem of finding the minimum size subset of these disks that still achieves this separation. We give a constant factor approximation algorithm for this problem.

We consider the scheduling problem of minimizing the average weighted completion time on identical parallel machines when jobs are arriving over time. For both the preemptive and the nonpreemptive setting, we show that straightforward extensions of Smith's ratio rule yield smaller competitive ratios compared to the previously best-known deterministic on-line algorithms, which are (4+epsilon)-competitive in either case. Our preemptive algorithm is 2-competitive, which actually meets the competitive ratio of the currently best randomized on-line algorithm for this scenario. Our nonpreemptive algorithm has a competitive ratio of 3.28. Both results are characterized by a surprisingly simple analysis; moreover, the preemptive algorithm also works in the less clairvoyant environment in which only the ratio of weight to processing time of a job becomes known at its release date, but neither its actual weight nor its processing time. In the corresponding nonpreemptive situation, every on-line algorithm has an unbounded competitive ratio

Standard local search algorithms for combinatorial optimization problems repeatedly apply small changes to a current solution to improve the problem's given objective function. In contrast, non-oblivious local search algorithms are guided by an auxiliary potential function, which is distinct from the problem's objective. In this thesis, we compare the standard and non-oblivious approaches for a variety of problems, and derive new, improved non-oblivious local search algorithms for several problems in the area of constrained linear and monotone submodular maximization. First, we give a new, randomized approximation algorithm for maximizing a monotone submodular function subject to a matroid constraint. Our algorithm's approximation ratio matches both the known hardness of approximation bounds for the problem and the performance of the recent ``continuous greedy'' algorithm. Unlike the continuous greedy algorithm, our algorithm is straightforward and combinatorial. In the case that the monotone submodular function is a coverage function, we can obtain a further simplified, deterministic algorithm with improved running time. Moving beyond the case of single matroid constraints, we then consider general classes of set systems that capture problems that can be approximated well. While previous such classes have focused primarily on greedy algorithms, we give a new class that captures problems amenable to optimization by local search algorithms. We show that several combinatorial optimization problems can be placed in this class, and give a non-oblivious local search algorithm that delivers improved approximations for a variety of specific problems. In contrast, we show that standard local search algorithms give no improvement over known approximation results for these problems, even when allowed to search larger neighborhoods than their non-oblivious counterparts. Finally, we expand on these results by considering standard local search algorithms for constraint satisfaction problems. We develop conditions under which the approximation ratio of standard local search remains limited even for super-polynomial or exponential local neighborhoods. In the special case of MaxCut, we further show that a variety of techniques including random or greedy initialization, large neighborhoods, and best-improvement pivot rules cannot improve the approximation performance of standard local search.

This thesis is a survey on the approximability of the asymmetric traveling salesmanproblem with triangle inequality (ATSP).In the ATSP we are given a set of cities and a function that gives the cost of travelingbetween any pair of cities. The cost function must satisfy the triangle inequality, i.e.the cost of traveling from city A to city B cannot be larger than the cost of travelingfrom A to some other city C and then to B. However, we allow the cost function tobe asymmetric, i.e. the cost of traveling from city A to city B may not equal the costof traveling from B to A. The problem is then to find the cheapest tour that visit eachcity exactly once. This problem is NP-hard, and thus we are mainly interested in approximationalgorithms. We study the repeated cycle cover heuristic by Frieze et al. We alsostudy the Held-Karp heuristic, including the recent result by Asadpour et al. that givesa new upper bound on the integrality gap. Finally we present the result ofPapadimitriou and Vempala which shows that it is NP-hard to approximate the ATSP with a ratio better than 117/116.

The <i>mixed postman problem</i> consists of finding a minimum cost tour of a mixed graph <i>M</i> = (<i>V</i>,<i>E</i>,<i>A</i>) traversing all its edges and arcs at least once. We prove that two well-known linear programming relaxations of this problem are equivalent. The <i>extra cost</i> of a mixed postman tour <i>T</i> is the cost of <i>T</i> minus the cost of the edges and arcs of <i>M</i>. We prove that it is <i>NP</i>-hard to approximate the minimum extra cost of a mixed postman tour. A related problem, known as the <i>windy postman problem</i>, consists of finding a minimum cost tour of an undirected graph <i>G</i>=(<i>V</i>,<i>E</i>) traversing all its edges at least once, where the cost of an edge depends on the direction of traversal. We say that <i>G</i> is <i>windy postman perfect</i> if a certain <i>windy postman polyhedron O</i> (<i>G</i>) is integral. We prove that series-parallel undirected graphs are windy postman perfect, therefore solving a conjecture of Win. Given a mixed graph <i>M</i> = (<i>V</i>,<i>E</i>,<i>A</i>) and a subset <i>R</i> &#8838; <i>E</i> &#8746; <i>A</i>, we say that a mixed postman tour of <i>M</i> is <i>restricted</i> if it traverses the elements of <i>R</i> exactly once. The <i>restricted mixed postman problem</i> consists of finding a minimum cost restricted tour. We prove that this problem is <i>NP</i>-hard even if <i>R</i>=<i>A</i> and we restrict <i>M</i> to be planar, hence solving a conjecture of Veerasamy. We also prove that it is <i>NP</i>-complete to decide whether there exists a restricted tour even if <i>R</i>=<i>E</i> and we restrict <i>M</i> to be planar. The <i>edges postman problem</i> is the special case of the restricted mixed postman problem when <i>R</i>=<i>A</i>. We give a new class of valid inequalities for this problem. We introduce a relaxation of this problem, called the <i>b-join problem</i>, that can be solved in polynomial time. We give an algorithm which is simultaneously a 4/3-approximation algorithm for the edges postman problem, and a 2-approximation algorithm for the extra cost of a tour. The <i>arcs postman problem</i> is the special case of the restricted mixed postman problem when <i>R</i>=<i>E</i>. We introduce a class of necessary conditions for <i>M</i> to have an arcs postman tour, and we give a polynomial-time algorithm to decide whether one of these conditions holds. We give linear programming formulations of this problem for mixed graphs arising from windy postman perfect graphs, and mixed graphs whose arcs form a forest.